Background: Diabetic Retinopathy (DR) is a major cause of vision loss. artificial intelligence powered optical coherence tomography angiography (OCTA) provides a non-invasive, quantitative method to detect early retinal ischemic changes by analyzing foveal avascular zone (FAZ) remodeling.

Objective: This study utilizes AI-powered OCTA metrics to assess FAZ remodeling and macular vessel density, evaluating their predictive value for early diabetic retinopathy and their association with visual function.

Methods: This prospective cross-sectional study evaluated 100 participants (50 with early diabetic retinopathy, 50 healthy controls). Using AI-based OCTA, we quantified FAZ morphology and macular vessel density. Statistical analysis, including multivariate logistic regression, assessed microvascular remodeling and identified independent predictive biomarkers for early disease detection.

Results: Early DR cases exhibited significantly larger FAZ area (0.41 ± 0.12 vs 0.27 ± 0.07 mm², P-value below 0.001), increased perimeter (2.68 ± 0.34 vs 2.05 ± 0.21 mm, P-value below 0.001), and higher acircularity index (1.29 ± 0.11 vs 1.08 ± 0.06, P P-value below 0.001). Macular vessel density in both superficial and deep capillary plexuses was significantly reduced in the DR group. FAZ enlargement correlated negatively with vessel density (r = −0.58 and −0.61, respectively; P-value below 0.001), and morphological distortion (acircularity/perimeter) correlated negatively with best-corrected visual acuity. Logistic regression confirmed FAZ metrics and reduced vessel density as independent predictors of early DR.

Conclusion: AI-based OCTA identifies significant FAZ enlargement and vessel density reduction in early DR. These metrics, particularly the acircularity index, are vital biomarkers for early detection and risk stratification.